Ensemble Learning Integrated with Cancer Survivor Intervention Trials

• Activity Number: 341 - Random Effects and Mixed Models
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Anjishnu Banerjee* and Melinda Stolley and Avik Chakrabarti and Alexis Visotcky
• Companies: and Medical College of Wisconsin and University of Wisconsin Milwaukee and Medical College of Wisconsin
• Keywords: Graphical models; SEMS; Bayesian nonparametric; Cancer survivorship; Mixed effects; Functional data

Abstract

The U.S. cancer survivor population is comprised of over 15 million individuals and is expected to grow to 20 million by 2026. Cancer incidence projections show that highest increases are expected among older adults and minorities. Many factors drive these differences, among which obesity is an important contributor. Mounting evidence suggests that obesity, unhealthy diets, and sedentary lifestyles adversely impact survivors' quality of life and comorbid illnesses while contributing to cancer mortality and progression. We develop a novel class models, modeling mixed effects from repeated measurements from minority cancer survivors, which helps explain "overall" effects of interventions, rather than considering them in isolation. Modeling of these interdependences would not only lead to more robust statistical inference but it would also lead to improved understanding of the factors leading to better health outcomes in cancer survivors. Combined with econometric techniques, the outcome targets demonstrated in this project help to identify the most optimal strategies in modifying, scaling up and expanding the previously concluded lifestyle intervention trials for cancer survivors.